Often universities do this, they may own the patent and license it to the company or take a cut etc. Arizona State University appear to do this through Skysong Innovations:

https://skysonginnovations.com/startups/list/

It's interesting they got a lot of funding from over 100 families with autism children:

https://skysonginnovations.com/startups/list/

It's important to remember that there are many Monet paintings that critics don't like, or that aren't 'monet enough'. He painted fast to sell and make money and many think some paintings aren't as finished as they could be. He himself destroyed a number of water lily paintings before an exhibition [1], and again a lot of the work he did when he was partly blind due to cataracts.

[1] https://content.time.com/time/specials/packages/article/0,28...

But memory is a quite a specialist manufacturing process, they couldn't just send a design to TSMC and get the same quality and cost. It would take years (decades) to create their own factories that might be able to produce competitive memory. If they use a third party to manufacture with existing skills (Samsung, SK Hynix, Micron etc) they might as well just use their designs too (and buy their chips)

Static RAM needs 6 transistors per cell, an M2 Ultra has 134 billion transistors, do the division and you get... "it doesn't work that way".

Static RAM is not DRAM.

You certainly could do try a 20bn cell SRAM, in 155mm^2, if you could handle the routing, but the power consumption might surprise you.

"Decades" seems excessive

But I'm guessing that doesn't work with someone who's been collecting other logins:

> Muneeb had been assembling usernames and passwords—5,400 of them taken from his own company’s network data.

> I've seen submissions get 3 upvotes in the first minute and never another upvote.

Is that people looking at the new page and seeing/upvoting your post and then it falls off the front.

That is many people's posts when I am reviewing new submissions.

Now upvoted, but it was negative

You can see his terror on his face and the lip trembling at his sudden realization somethings gone wrong at 24 seconds:

https://www.youtube.com/watch?v=e6Y2uQn_wvc&t=24s

> But respect the community, and only share what is truly relevant. Save the crayon pictures for your kitchen fridge.

That highlights the problem - its not AI - it's the oversharing thats the issue. Many people have moved from "Sharing whats unusual/interested/excited me" to "What can I share today".

The constant stream of mediocrity drove me away from Facebook (years ago) and then Instagram.

Related: March EV sales up 51% in mainland Europe https://www.theguardian.com/environment/2026/apr/20/electric...

I think the big difference is that criminal wire fraud depends on a "clear scheme to defraud with intent". Tesla/Musk can argue that they thought they would delivery - They've been making claims that FSD was coming for years and have been slowly making deliveries towards FSD, its just that its harder/taken longer than expected and without a smoking gun (email chains like in the Holmes case) it would be very hard to prove.

They may have committed false advertising or "failed to deliver on contract" but they are civil matters, which could still involve big payouts, but not prison time.

There's a corpus of work that could help there. Tesla was forced to add disclaimers like "Elon's statements are aspirational and do not necessarily represent engineering reality", as well as quotes from him on investor calls where he's described (in 2009, I believe) FSD as a "solved problem, we're just implementing", and five years later, "Our highest priority is solving the problem of FSD". But it seems possible that there comes a time when an ambitious prosecuting attorney or attorney general pushes for this and the discovery that comes with (though I have near zero confidence that even then, that discovery won't already be thoroughly crippled by document retention policies or outright fuckery by Elon).